Previously,Iwrote"Hands-onDeepLearning",whichwasanintroductoryclassicinthecircle.Later,ItaughtStanfordAIcoursesforfreeatStationB.Avideoofadifficultandhard-coredissertationwasplayed360,000.Manyresearchgroupswentfromtutorstoundergraduateapprentices.Allmylifeischasingfan.
Withsuchgreatsharing,it’snowondermanypeopleproudlyclaimto"havebeenworshippedbytheteacher"——
Recently,Idiscoveredthatsomeplatformshavedonesomethingmoreexciting.
Nottomentionthatthisplatformprovidesfreecomputingpower,italsoputstheintroductoryclassic"hands-ondeeplearning"intotheplatform,freeforeveryonetolearnandpractice,called:AmazonSageMakerStudioLab.
Accordingtotheintroduction,theplatformisbasedonJupyterLabandprovidesfreeGPUandCPUcomputingpower+15Gpermanentstorage.ItisalsolinkedtoGitHubandsupportstheuseofmainstreammachinelearningtoolcomponentsandopensourceresourcepacks.Developerscancombinethe"Hands-onDeepLearning"textbookTrainthemodelyourselfandseetheresults.
Moreover,theyclaimthattheyonlyneedanemailaddress,noofficialaccount,andnocreditcard.
istheresuchagoodthing
willtakeeveryonetotestittogethertoday.
Bylinkingstudiolab.sagemaker.aws,wecanlogintothewebplatformtoexplorethereality.
Asyoucanseefromtheinterface,theplatformprovidesGPU/CPUcomputingpoweroptions,andcanbeuseddirectlywithoutpayment.
Inthelowerrightcorner,wecanseetheDiveintoDeepLearning(abbreviatedasD2L).
ClickOpenD2Lnotebooksdirectlytoopen:
Afterstartingtheproject,thesystemautomaticallyloadsD2Lresourcesandstorestheminourcloudfolder.
TheREADMEfilehasalsobeenopened.Here,thesystemenvironmentconfiguration,thesummaryofthebook,theaudience-oriented,andthecatalogframeworkareallavailable.Therearelinkstoeachchapterattheendofthearticle,fromwhichyoucandirectlyenter.
Atthispoint,youcanlearndeeplearningthroughtheplatformcombinedwithcoursecontentandpracticaloperations——
Sowhatistheactualeffect
TaketheclassicAlexNetpartoftheclassicsasanexampletogetafeelforit.
Inordertoensureabetterunderstandingforbeginners,wecanalsobuildasingle-channeldataexampletoobservetheoutputofthe8layersinAlexNet.Thepurposeistohelpusintuitivelyunderstandtheroleofdifferentlayers:
Themostimportantthingisthatthetrainingmodellinkisalsoopenforpracticaloperation,justselectthecodeparttorun.
Seeingthatthecurveisdrawnslowly
Itisworthmentioningthat,becausetheentireinterfacecanbeusedasanotebooktoaddcodetorecordlearningandthinking——
Therefore,evenifourcourseiscompleted,wecanaddacodecolumnattheendofthetext,andprogramthehomeworkbasedontheexercisesattheendofthechapter.
Theabovedisplayisjustasubsection.Infact,fromtheintroductionofconceptssuchasfullyconnectedlayer,convolution,andpooling,totheexplanationofResNetandDenseNet...theyarepresentedandoperatedinthefreeenvironmentofAmazonSageMakerStudioLab,andtheyareallarrangedclearly.
Theplatformalsocarefullyconsidersthatourhigh-levellinearalgebrahasdifferentfoundations.Noteveryonehaspassed90points.Italsogivesanintroductionandcodeimplementationofsingle-variablecalculus,maximumlikelihoodandothermathematicalfoundations,andalsoattachesit.Theenvironmentconfigurationmethodisaveryreliableposture.
Aftertheaboveverification,thisAmazonSageMakerStudioLabcanindeedimplementthe"hands-ondeeplearning"ofthegreatgodforfreeandcomplete——
ForthosewhowanttogetstartedorevenhaveadeepgraspoftheAI/MLtechnology,thistheory+practicestudymethodisnaturallymoreefficient,andthetransitionwillbemoreseamlesswhenswitchingtoactualworkorscientificresearchorevenentrepreneurshipinthefuture.
Infact,itsabilityisnotonlyatthislevel.
Afreeplatformfordevelopers
Fromthename,youalsodiscoveredthatthecompanybehindthelaunchofAmazonSageMakerStudioLabisAmazonCloudTechnology.
Thiscutting-edgetechnologymanufacturerlaunchedafreeplatformthistime,notonlyturning"hands-ondeeplearning"intoatheory+practicalexercisefield,butalsowantstofacedatascientists,enterprisedevelopers,universityteachersandstudents——
providesafreelow-thresholdentrymachinelearningopportunityinclusiveofbenefits.
Infact,beforeAmazonCloudTechnology,thereweremanyopenmachinelearningplatformsintheindustry——
So,whatarethenewhighlightsofAmazonSageMakerStudioLabthistime
Theplatformprovides15Gormorepermanentstorage,16Gmemory,4CPUs,andtheGPUisNVIDIATeslaT4,whichisslightlyhigherthanothermainstreamplatforms.
DuetotheuseofthenewerarchitectureofNvidiaTeslaT4,itsmixed-precisioncomputingspeedindicatorsarecorrespondinglyhigher.Inaddition,thefreeversionusesthesamearchitectureasSageMakerStudio,whichisequivalenttoalayerofenterprise-levelBuff,andthestabilityismoreguaranteed.
Itisworthnotingthattheplatformadvertises4hoursGPU+12hoursCPU,butinfact,wecanstillopenRuntimeagainafterthetimeisup,andtheoriginalfilesstillexist.
Butifyouwanttominecoins,forgetit...
TheplatformexpresslyprohibitstheuseofSageMakerStudioLabforproductionactivities,andthedirecttitleisfoundbyminingencryptedcurrency.
Intermsofoperability,AmazonSageMakerStudioLabismoreconciseandintuitivethanotherplatforms.
TheinterfacecannotonlycreateJupyterNotebookfiles,butalsosupportsustodirectlycreateTerminaltabsandMarkdownformatfiles.
Inaddition,thisplatformisequippedwithCondaandPipresourcepackagemanagerstoavoidrepeatedinstallationofopensourcesoftwarepackages,savingtroubleandworry.
Ifthereisaymlenvironmentconfigurationfileintheclonedproject,whentheprojectiscreated,theCondaenvironmentwillalsobeestablishedsynchronously.
TheplatformisalsolinkedtoGithub.
AddthefollowingcontenttoourownGithubprojectREADMEdocument:
Youcanaddthe"OpeninStudioLab"button——
OtherscanaccesstheAmazonSageMakerStudioLabplatformwithjustoneclick.
Ofcourse,wecanalsocreateanewipynbbyuploadingorcopyingmanually.
Taketheclassicimageclassificationalgorithmasanexample.
Aftercopyinganipynbfilefromanotherplatform,itcanbeusedwithoutmodification.Atmost,itmeansmanuallyinstallingdependencies.
Actualtrainingisbasicallythesameasonotherplatforms,andsometimes,evenslightlyfaster.
Finally,wehavetolookatdatasecurityissuesthatmanypeoplecareabout.
Wedomachinelearningandoftenholdalargeamountofprivatefacialinformation,evenhospitalpatientinformationthathasnotbeendesensitized.Inordertoprotecttheprivacyanddatasecurityofothers,wehavetolookatthisaspect.
Atpresent,manyfamousschoolsandenterpriseshaveusedAmazonSageMakerStudioLabandendorsedtheirsites.
Amongthem,thereistheSchoolofEngineeringoftheUniversityofPennsylvania,whereENIACwasborn,theDepartmentofFinance,SantaClaraUniversity,California,andHuggingFace.
Therearealsomanydomesticfollowers.
Adoctoralstudentinthefieldofmachinelearningfroma985PolytechnicUniversityinthesouthsaidthatalthoughthedirectionoftheirresearchgroupistraditionalmachinelearning,itstillneedsdeeplearningtoassistinverification.
Sincethelaboratory’scomputingequipmentwaspurchasedseveralyearsago,withtheincreaseinpersonnelandresearchdirections,especiallyontheeveofsubmission,competitionforcomputingresourcesiscommon.AmazonSageMakerStudioLabisindeedattractivetothem.
Aftertalkingabouttheadvantagesofthisfreeplatform,thenextquestionis:howtoapply
NoneedforanAmazonCloudTechnologyaccount,justlogintotheofficiallinkstudiolab.sagemaker.aws/requestAccountandfillintheemailandrelatedinformation.
However,inordertoensurethateveryonecanuseitthroughtheapplicationassoonaspossible,therearesomeTips,Ihopeyouwillpayattentionto:
ItisrecommendedthatthelanguagebeusedinEnglish,andthenameoftherelevantorganizationshouldbeclearlyfilledin,andthesuffixofthemailboxshouldmatchtheEnglishnameoftheorganization.Thiskindofapplicationismorecredibleandreliable.
Ifyoumeettheaboveconditions,youcangetaninvitationwithin24hoursofthepro-test.Payattentiontocheckyouremail.
Asmentionedearlier,AmazonSageMakerStudioLabandtheprofessionalversionofAmazonSageMakerStudiohavethesamearchitecture,soifyouwanttomigratefromthebeginnerfreeversiontotheprofessionalversion,itiscertainlynotabigdeal.
Forprofessionaldevelopers,thiscertainlyprovidesmoreadvancedresearchandentrepreneurialpossibilities.
Morespecifically,theprofessionalversionofAmazonSageMakerStudiohasprovideddeveloperswithacomprehensivesetoffeaturesfromstarttofinish:
Forexample,itprovideslarge-scaledistributedtrainingforourproductivelarge-scalemodeltrainingneeds.UsepartitioningalgorithmstoautomaticallysplitlargemodelsanddatasetsinGPUinstancestoimproveparallelismandspeeduptraining.
Forexample,thedatalabelingfunctionGroundTruthPlushasbroughtinhumanexperts,combinedwithmachinelearningtoassistpre-labeling,greatlyreducinglabelingerrorsandincreasingthelabelingrate.
AnotherexampleisAmazonSageMakerDataWrangler.Thisfunctionisorientedtothedatapreparationstageinmachinelearning,anddataselection,cleaning,andexplorationcanbeperformedthroughavisualinterface.Youonlyneedone-clickimport,andyoucanquicklystandardizeandconvertlargequantitiesofdatawithavarietyofstructureswithoutcode.Deloitte,oneofthe"BigFour",adoptedthisfunction.Datapreparationthatwasoriginallycompletedinafewmonthsisnowcompressedtoafewdays.
Inaddition,AmazonSageMakerStudioalsoincludesaccesscontrolmanagement,modelmonitoring,serverlessreasoningfunctions,reasoningconfigurationrecommendations...untilthefullcycleofindustrializedAI/MLservicesareallroundedup.
Forprofessionaldevelopersanddatascientists,whetheritisresearchorentrepreneurship,thesefeaturesprovidemorepossibilities.
Themachinelearningmarathonprojectisamanifestation.
Thisactivityisseveraltimesayear,andtheplatformwillpresenttestquestionsintheapplicationofAI-relatedfields,coveringAIautomationprogramming,disasterpreventionordamageassessment,etc.
Therearealotofactivitieslikethisthataregearedtoactualneedsandbenefitdevelopersfromtechnology,andeachhasitsownfunorsocialvalue.
AmazonDeepRacer,aself-drivingracinggamewithazerothresholdtogetstartedwithmachinelearning,hasmillionsoffollowersand140,000participatingdevelopers;
Therearealsoactivitiestocooperatewiththenon-profitorganizationGirlsinTechtohelpmorewomenunderstandandlearnonmobiledevices,andeliminatethegendergapinthetechnologycircle;
ThereisalsothelatestreleaseofAmazonSageMakerCanvans,whichhasattractedhighattentioninthecircle,andisaimedatinternalanalystsandoperatorswith0codeexperiencetohelpthemapplythetechnologyofmachinelearningintheiractualbusiness.
Fromabusinessperspective,thesepracticesarecertainlybeneficialtotheecologicalconstructionofthehome,andtheyarenecessaryactionsforleadingcompaniestomaintaintheirposition.Inaddition,alargenumberofcompaniesoftenrecruit500,000relatedpractitionerswithanannualsalaryof500,000,whichisalsoagoodthingformanydevelopers.
Afterall,theshortageofartificialintelligenceisvisibletothenakedeye.Mostdevelopershaverichprogrammingexperienceandmathematicalfoundations.Theonlybarrieristhattheyarenotfamiliarwithmachinelearning.Makeupthispieceanddrinkthesoup,whynotdoit
However,fromtheperspectiveoftheindustry,theaboveactionsareindeedpromotingthelandingofcutting-edgetechnologies——
Butthepushhereisnottobefaster,buttobewider.
Youknow,theautomobilewasjustinventedahundredyearsago,andonlymechanicalexpertscanbecomecarowners,sothatinsocietyatthattime,driverswereaprofessionwithcutting-edgetechnologyintheirhands.
Youmustalsoknowthat30yearsago,thePCandtheInternetwereonlytoysforasmallnumberofdevelopers,sothatthedevelopmentofawebsitecanquicklymakepeoplerichandaccumulateassetsashighastallbuildings.
Therefore,ahundredyearsago,peoplecouldnotunderstandasocietywhereeveryonecoulddrive.Therefore,30yearsago,itwasalsodifficultforpeopletoimaginethattheycouldmovetheirfingersandhavetheirownInternetplatform.
Thesamegoesformachinelearningtoday.Eventhemostcutting-edgetechnologymasterscanonlyglimpseasmallpartoftheAIfloorplan.Onlybyrepeateddimensionalityreduction,thistechnologycanenterthesceneofthousandsofindustries,anditwillproducedifferentfrequenciesofreverberationinthehandsofpeoplewithdifferentbackgroundsanddifferentexperiences——
So,howmuchenergycanAItechnologyreleaseinthefuture
answerhastobefoundineveryindividualdeveloperandeveryscenerift.
Formoretechnicalwinddirections,longpressthepicturebelowtounderstand:
Unconsciously,bytheendoftheyear,forthetechnologycircle,thedevelopmentofvarioustechnologiesandindustriesduringtheyearhasnotonlystoodatthepeak,butalsoexperiencedupsanddowns.
Atthelastmomentof2021,AmazonCloudTechnologywantstolistentothevoicesofdevelopersinthecloudcomputingfield.Forthisreason,theaward-winningsurveyofcloudcomputingdevelopershasofficiallyopened.Wesincerelyinviteallpartnerstoparticipate,andmultiplegiftsarewaitingforyou!
思否编辑部官方账号，欢迎私信投稿、提供线索、沟通反馈。